Reservoir bank landslide deformation mechanism analysis method in combination with InSAR deformation characteristics and fluid-structure interaction

Abstract

The invention discloses a reservoir bank landslide deformation mechanism analysis method in combination with InSAR deformation characteristics and fluid-solid coupling, and the analysis method comprises the steps: firstly, recovering a deformation field of a reservoir bank landslide through employing a multi-temporal InSAR, and carrying out the induction factor analysis of a deformation independent component in combination with an ICA method; secondly, the saturated permeability coefficient of the slip mass material is corrected in an auxiliary mode based on the InSAR deformation time sequence; then, a seepage analysis result is coupled to an FLAC3D model for stress-strain analysis; and finally, revealing the deformation mechanism of the reservoir bank landslide based on InSAR deformation spatio-temporal evolution characteristics and a fluid-solid coupling analysis result. According to the reservoir bank landslide deformation mechanism analysis method, reservoir bank landslide InSAR deformation spatio-temporal evolution, one-dimensional pore water diffusion model auxiliary correction saturated permeability coefficient and coupling pore pressure stress analysis are carried out aiming at the limitation of InSAR and a numerical simulation technology in the reservoir bank landslide deformation problem.

Description

Technical Field

[0001] The invention relates to the technical field of reservoir bank landslide deformation mechanism analysis, and in particular to a reservoir bank landslide deformation mechanism analysis method combining InSAR deformation characteristics and fluid-solid coupling. Background Art

[0002] The construction of large reservoirs has brought significant social and economic benefits in flood control, power generation, shipping, irrigation and tourism. However, the impoundment of reservoirs easily causes strain softening of the rock and soil of the reservoir bank, which is the direct cause of the rebirth of reservoir bank landslides and the revival of ancient landslides. Such reservoir bank landslides are widely found in the reservoir areas of many world-class water conservancy and hydropower projects, such as Pubugou, Laxiwa, Baihetan, etc. Since rainfall and water level fluctuations can directly affect the stability of landslides, these landslides pose a serious threat to reservoir safety, such as swells and river blockages. Therefore, the problem of reservoir bank landslides has always received special attention from researchers.

[0003] Surface deformation is a direct manifestation of the movement of reservoir bank landslide blocks. It is well known that this movement is affected by internal and external factors such as rainfall, water level fluctuations, and its own geological structure. For a long time, many researchers have used deformation monitoring to understand the spatiotemporal evolution characteristics of landslide movement, which is very necessary for landslide prevention and early warning. GNSS is a conventional monitoring method. Although it has high accuracy, sparse monitoring points make it difficult to describe the spatial deformation pattern of reservoir bank landslides. The synthetic aperture radar interferometry (InSAR) technology developed in recent decades has become an important means for early identification and deformation monitoring of reservoir bank landslides due to its advantages of high spatiotemporal resolution and the ability to trace historical deformation. In addition, the dense monitoring points of InSAR technology provide a rich data set for analyzing the correlation between deformation signals and triggering factors. Many researchers have studied the triggering mechanism of reservoir bank landslides through cross-correlation analysis of deformation response characteristics. Shi xuguo et al. (2021) revealed a high correlation between the displacement of the Guobu landslide and rainfall through InSAR results during the stable operation of the Laxiwa Hydropower Station, and believed that the deformation was mainly controlled by rainfall. Zheng wanji et al. (2023) decomposed the InSAR spatiotemporal deformation signals by independent component analysis (ICA) and calculated the correlation coefficients between these signals and rainfall and water level fluctuations to reveal the triggering mechanism and time delay of the Xinpu landslide complex. This correlation analysis estimates the lag time of the deformation response to different triggering factors, making it possible to discuss the deformation evolution characteristics of complex reservoir bank landslides. However, InSAR results can only qualitatively interpret the response relationship of deformation from a geometric perspective, and it is difficult to describe the stress-strain relationship between deformation and triggering factors and reveal the potential slip surface characteristics. In addition, the deformation scales of different landslides are significantly different, and it is difficult to form a unified threshold to assess the instability risk of landslides. However, this risk assessment is very necessary to assist in the prevention and early warning of landslide disasters.

[0004] Unlike surface displacement monitoring, numerical simulation technology is based on the constitutive model of geotechnical materials and combines internal and external dynamic boundary conditions to quantitatively solve the seepage field, stress-strain field and stability coefficient of reservoir bank landslide. It is well known that changes in groundwater conditions are the main cause of deformation or instability of reservoir bank landslide. Many scholars have proposed practical empirical formulas to fit soil-water characteristic curves and permeability coefficient functions (for example: Gardner, Brooks-Corey, Van Genuchten, Fredlund-Xing models), which are helpful for reasonable unsaturated soil seepage analysis of reservoir bank landslide. In addition, based on the effective stress principle, Fredlund (1978) developed the shear strength theory of unsaturated soil and derived the constitutive equation of unsaturated soil media together with Rahardjo (1993). This equation can quantitatively solve the stress state of reservoir bank landslide under coupled dynamic seepage conditions. We can use the relationship between the anti-sliding force and the sliding force on the sliding body to evaluate the stability of reservoir bank landslide. In addition, the strength reduction method can also use this stress state to simulate and calculate the shape of the potential slip surface and possible failure deformation mode. These studies have made numerical simulation methods an important means to reveal the deformation mechanism, instability risk and failure mode of reservoir bank landslides. However, due to the complexity of the three-dimensional geological structure, characteristic profiles are usually selected for research. Obviously, this is not enough to analyze the spatial evolution characteristics of landslide deformation. In addition, in reservoir bank landslides, the rationality of numerical simulation results is easily restricted by the accuracy of the saturated permeability coefficient. Due to the spatial heterogeneity and disturbance of sample data, it is very difficult to accurately determine the saturated permeability.

[0005] Many scholars have conducted corresponding research on typical reservoir bank landslides, whether using InSAR deformation monitoring or numerical simulation. However, both methods have certain limitations in the study of reservoir bank landslide deformation. Summary of the invention

[0006] In view of the limitations of InSAR and numerical simulation technology in the problem of reservoir bank landslide deformation, the present invention provides a reservoir bank landslide deformation mechanism analysis method combining InSAR deformation characteristics and fluid-solid coupling.

[0007] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0008] A deformation mechanism analysis method for reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling. The analysis method firstly uses multi-temporal InSAR to restore the deformation field of the reservoir bank landslide and combines the ICA method to analyze the inducing factors of the independent components of the deformation; then, the saturated permeability coefficient of the sliding material is corrected based on the InSAR deformation time series; then, the seepage analysis results are coupled to the FLAC3D model for stress-strain analysis; finally, the deformation mechanism of the reservoir bank landslide is revealed based on the InSAR spatiotemporal evolution characteristics and fluid-solid coupling analysis results.

[0009] As a further improvement of the above technical solution:

[0010] Preferably, the analysis method comprises the following steps:

[0011] Step S1, acquiring SAR, hydrological and geological data: hydrological data is used to analyze InSAR deformation driving factors and time delays and as boundary conditions for numerical simulations, and geological data is used to establish numerical models;

[0012] Step S2, TCP-InSAR data processing: using the temporary coherence point InSAR method to restore the InSAR deformation field of the reservoir bank landslide;

[0013] Step S3, analysis of the spatiotemporal evolution of InSAR deformation and time delay of inducing factors: independent component analysis based on the independence assumption decomposes the InSAR deformation into multiple components with different spatial scores and temporal characteristics; the decomposed independent components are used as the spatiotemporal evolution of deformation induced by different driving factors, and the correlation coefficient between the deformation response and the triggering factors is evaluated by cross-correlation analysis to analyze the driving factors and time delay of the reservoir bank landslide.

[0014] Step S4, fluid-solid coupling numerical model: establish a multi-field information fluid-solid coupling numerical model of the reservoir bank landslide that can be quickly converted between GeoStudio and FLAC3D, and conduct flow stability analysis of the reservoir bank landslide under conditions of rainfall and water level fluctuation;

[0015] Step S5, deformation mechanism of reservoir bank landslide: using the driving factors of InSAR deformation as hydraulic boundary conditions, stress-strain analysis is performed to analyze the failure deformation mode, stress-strain state and stability of different inducing factors.

[0016] Preferably, the step S3 includes the following steps:

[0017] S3-1, select the number of independent components: use principal component analysis to reduce dimension and noise to select independent components, and the variance ratio of each component is used as an empirical indicator for selecting principal components;

[0018] S3-2, ICA decomposition: ICA decomposes the mixed observation signals on the SAR time series into linear combinations of independent components with statistical characteristics; the temporal deformation of each pixel is the sum of independent deformation signal sources and noise; the surface deformation signals and errors caused by different driving factors are extracted and decomposed;

[0019] S3-3, InSAR deformation spatiotemporal evolution and correlation analysis: Combined with spatial evolution, the time vector of the independent component is correlated with reservoir water level fluctuation, accumulated rainfall and reservoir water joint effect to analyze the potential driving factors.

[0020] Preferably, in the ICA decomposition, a set of InSAR observations X is set 1 (t),X 2 (t),…,X n (t), t is the observation time, and the observation value is generated by the linear combination of independent components, which is:

[0021] [X 1 (t),X 2 (t),…,X n (t)] T =A.[S 1 (t),S 2 (t),…,S n (t)] T (6)

[0022] Among them, A is an unknown coefficient matrix representing the linear combination of different components; S(t) is an independent component, including linear deformation, periodic deformation, step deformation, atmospheric delay and noise error; when X is known, i (t) Estimate the coefficient matrix A and independent components S i (t).

[0023] Preferably, in the InSAR deformation spatiotemporal evolution, FastICA is used to process the InSAR time series, and the central observation matrix X is first calculated according to the observation matrix X C :

[0024]

[0025] in, Each column of is equal to the average of the corresponding column in X, and the center matrix X C Each column of has a mean of zero; the covariance matrix C is generated using the center matrix X for:

[0026]

[0027] The covariance matrix is ​​decomposed by maximizing the total variance of the projection based on principal component analysis:

[0028]

[0029] Among them, e and E are eigenvectors and eigenvector matrices, d and D are eigenvalues ​​and eigenvalue matrices respectively; the calculation formula of the whitening matrix is:

[0030] Q=(ED -1 / 2 E T ) k (10)

[0031] Among them, k is the number of principal components retained, and the observation matrix Z after whitening is written as:

[0032] Z=QX C (11)

[0033] After the observation matrix is ​​centered and whitened, the FastICA algorithm estimates the source signal matrix S and the mixing matrix A by maximizing the spatial non-Gaussianity:

[0034]

[0035] in, and are the mixing matrix and source matrix obtained from the center and whitened observations respectively. By using the inverse operation of the whitening process, the center observation matrix is ​​expressed as a combination of the source matrix and the mixing matrix:

[0036]

[0037] Among them, Q + is the pseudo-inverse of the whitening matrix Q; the original observation matrix is:

[0038]

[0039] By performing FastICA processing on TCP-InSAR observations, the temporal eigenvectors and spatial scores of independent components are separated.

[0040] Preferably, the step S4 includes the following steps:

[0041] S4-1, InSAR deformation inversion of saturated permeability: Invert the hydraulic diffusion coefficient based on the rainfall-induced temporal deformation decomposed by ICA and the daily rainfall records, and obtain the saturated permeability estimate based on the relationship between the hydraulic diffusion coefficient and the permeability coefficient; estimate the soil-water characteristic curve and the permeability coefficient function in combination with the geotechnical test results;

[0042] S4-2, Geostudio unsaturated seepage field: hydraulic boundary conditions were set in Geostudio unsaturated seepage analysis, and a numerical model of Geostudio unsaturated seepage analysis of reservoir bank landslide was established;

[0043] S4-3, quick conversion between Geostudio and FLAC3D: the finite element mesh model is stretched in the normal direction, and the original triangle and quadrilateral units in the stretched three-dimensional model are changed into wedge and hexagonal block units; the corresponding relationship is established according to the node numbering rules of different mesh types of the two-dimensional model and the three-dimensional model; the pore water pressure, saturation and volume water content in the finite element seepage field are converted to the FLAC3D model in a 1:1 ratio;

[0044] S4-4, initial stress and parameter correction: In the fluid-solid coupling model, based on the effective stress principle and the relationship between pore water pressure, matrix suction, total stress and saturation, the seepage information is extracted to correct the initial stress of the saturated and unsaturated zones; and the geotechnical material parameters are zoned and corrected based on the results of the seepage field.

[0045] Preferably, the inversion process of the hydraulic diffusion coefficient is:

[0046] The formula for expressing the transient pore water pressure change caused by rainfall infiltration is:

[0047]

[0048] Where P is the pore water pressure, D is the hydraulic diffusion coefficient, and z is the depth. Formula (15) describes the propagation process of pore water pressure inside the landslide. The rainfall record R(t) and the scale factor q are used to describe the pore water pressure on the surface:

[0049] P(t,z=0)=qR(t) (16)

[0050] The analytical solution of the propagation process is:

[0051]

[0052] Where s is the time variable; the hydraulic diffusion coefficient and permeability coefficient are combined as follows:

[0053]

[0054] Where C is a measure of the change in volumetric water content with pore pressure, and takes the minimum value C when the soil is saturated. 0 , the hydraulic diffusion coefficient reaches a peak value D 0 , k s represents the saturated permeability coefficient.

[0055] Preferably, the correction method of the initial stress is:

[0056] According to the effective stress principle, the effective stress σ′ of rock and soil mass is expressed as:

[0057]

[0058] Where σ is the total stress, u a is the pore gas pressure; σ s is the suction stress, which is equal to the pore pressure when the soil is saturated and is taken as the matrix suction when the soil is unsaturated. r is saturation;

[0059] The suction stress and total stress of the model before the pore water pressure is introduced are set to be σ s 0 and σ 0 , the suction stress of the model after import is σ s 1 , the corrected initial total stress σ 1 It is expressed as:

[0060] σ 1 =σ 0 -(σ s 0 -σ s 1 ) (twenty one).

[0061] Preferably, the parameter correction method is:

[0062] Let ρ d is the dry density of the material, using the porosity n and the density of water ρ w Correction for the natural density ρ:

[0063] ρ=ρ d +nS r ρ w (twenty two)

[0064] For the unsaturated zone, the effective cohesion is modified using the pore water pressure and the friction angle:

[0065]

[0066] Where C is the cohesion under saturation state.

[0067] Preferably, the step S5 includes the following steps:

[0068] S5-1, stress boundary conditions: setting boundary conditions to constrain the numerical model;

[0069] S5-2, calculation condition setting: according to the different stages of the landslide and different conditions of external dynamics, the working condition setting is carried out, and the mechanical mechanism of landslide deformation is analyzed based on the stability, damage deformation characteristics and mechanical state under different scenarios;

[0070] S5-3, Dynamic response of reservoir bank landslide stability during InSAR observation period: The water level fluctuation and daily rainfall intensity during the InSAR observation period were used as calculation conditions to observe the spatiotemporal evolution of simulated deformation and the dynamic response of stability.

[0071] Compared with the prior art, the reservoir bank landslide deformation mechanism analysis method provided by the present invention, which combines InSAR deformation characteristics and fluid-solid coupling, has the following advantages:

[0072] The reservoir bank landslide deformation mechanism analysis method combining InSAR deformation characteristics and fluid-solid coupling of the present invention analyzes the deformation mechanism of the reservoir bank landslide based on a new perspective of InSAR and numerical simulation technology: the TCP-InSAR technology combined with the ICA method can quickly obtain the refined deformation spatiotemporal evolution characteristics of the reservoir bank landslide and reveal potential inducing factors; based on this, the one-dimensional pore water diffusion model can be used to assist in correcting the permeability coefficient of geotechnical materials, which is beneficial to improving the rationality of the unsaturated seepage analysis of the reservoir bank landslide; Geostudio and FLAC3D models are quickly converted to establish a fluid-solid coupling numerical analysis model for revealing the destruction deformation mode, stability and stress state of the reservoir bank landslide; based on the InSAR deformation evolution characteristics and fluid-solid coupling stress state, the deformation mechanism of the reservoir bank landslide is assisted in revealing. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is the overall flow chart of the present invention.

[0074] Figure 2 It is a schematic diagram of the numerical model of the seepage field and stress-strain field of the present invention.

[0075] Figure 3 It is a schematic diagram of the conversion between Geostudio and FLCA3D seepage field model of the present invention.

[0076] Figure 4 It is the InSAR average deformation rate result of a typical reservoir bank landslide in a reservoir area of ​​the present invention.

[0077] Figure 5 It is the result of InSAR deformation temporal and spatial evolution of a typical reservoir bank landslide in a reservoir area of ​​the present invention. DETAILED DESCRIPTION

[0078] The specific embodiments of the present invention are described in detail below. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0079] like Figure 1 As shown in the figure, the present invention proposes a deformation mechanism analysis method for reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling, which is mainly analyzed from three aspects, including the spatiotemporal evolution of InSAR deformation of reservoir bank landslide, the one-dimensional pore water diffusion model assisted correction of saturated permeability coefficient and coupled pore pressure stress analysis. Firstly, the deformation field of reservoir bank landslide is restored by multi-phase InSAR and the inducing factors of deformation independent components are analyzed by ICA method; then, the saturated permeability coefficient of sliding material is corrected based on InSAR deformation time series; then, the seepage analysis results are coupled to the FLAC3D model for stress-strain analysis; finally, the deformation mechanism of reservoir bank landslide is studied based on the spatiotemporal evolution characteristics of InSAR and the results of fluid-solid coupling analysis.

[0080] The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling of the present invention specifically comprises the following steps:

[0081] Step S1, obtaining SAR, hydrological and geological data

[0082] Sentinel-1A / B is a radar satellite launched by the European Space Agency (ESA) in 2014 to provide earth observation data for the Global Monitoring for Environment and Security (GMES) program. It is still operating stably in orbit. The satellite's products adopt an open access strategy and cover most areas of the world (https: / / sentinel.esa.int / web / sentinel / home). In IW imaging mode, the range resolution of Sentinel-1SAR images is about 2.3m, the azimuth resolution is about 13.9m, and the revisit period is 12d. At least one available orbital data in the ascending and descending orbits is covered throughout China. Therefore, this data product can be given priority for InSAR deformation monitoring in the reservoir bank landslide scenario. In terms of commercial data, ALOS-2PARSAR L and LT-1L are both long-waveband high-resolution satellites, which have strong decoherence suppression performance in reservoir bank landslide areas with dense vegetation coverage. In addition, the rain and water monitoring of most large reservoirs can provide historical water level data with a time resolution of 1d. The weather station can provide daily precipitation records with a time interval of 1 day. These hydrological data can be used to analyze the driving factors and time delays of InSAR deformation and as boundary conditions for numerical simulation. SRTMDEM with a resolution of 30m can be used for geocoding of SAR images and terrain phase removal. Typical reservoir bank landslides have been geologically surveyed, and engineering geological planes, profiles, geotechnical material parameters, etc. can be used to establish numerical models.

[0083] Step S2, TCP-InSAR data processing

[0084] InSAR deformation time series can be used to analyze the spatiotemporal evolution characteristics of landslide movement. However, the inversion of multi-time series SAR image deformation phase time series relies on MT-InSAR (commonly used MT-InSARs include PS-InSAR and SBAS-InSAR) methods to effectively suppress DEM errors, atmospheric delays, and incoherence noise. Although PS-InSAR and SBAS-InSAR have their own advantages in time series deformation processing, they both have certain limitations for reservoir bank landslide scenes. For example, dense vegetation makes it difficult for PS-InSAR technology to select enough PS points; high vegetation coverage and steep terrain make the SBAS-InSAR method inevitably introduce phase unwrapping errors.

[0085] The present invention uses a temporary coherent point InSAR (TCP-InSAR) method to restore the InSAR time-series deformation of the reservoir bank landslide.

[0086] TCP-InSAR data processing includes the following steps:

[0087] S2-1, SAR image registration.

[0088] Since the geometric deviation of time series SAR images can easily interfere with the quality of the interferometric phase, the SAR images need to be registered before MT-InSAR processing. Usually, a single master image is selected and combined with an external DEM for registration processing. The registration accuracy of Sentinel-1 data must be better than 0.00001 pixel.

[0089] S2-2, small baseline set interference pattern combination.

[0090] Using multiple interference pairs to form a small baseline set and constructing a deformation model from a time series to estimate deformation parameters is the core step of MT-InSAR processing. Considering that the terrain residual phase is proportional to the spatial baseline, a certain spatial baseline threshold is set in the interference pattern to suppress the influence of terrain residuals. The time interval of the interference pair is an important cause of decoherence, and a short time baseline can suppress decoherence noise. The interference pattern combination strategy is the basis for deformation estimation and separation of atmospheric delays. However, multiple baseline subsets will cause rank deficiency in deformation inversion. Therefore, in the present invention, the interference pattern is combined with a single subset.

[0091] S2-3, D-InSAR processing.

[0092] The SAR images with short temporal and spatial baselines are interferometrically processed to obtain the interferometric phase map. Considering the difference between the radar coordinate system and the geographic coordinate system, an external DEM (SRTM DEM) with a resolution of 30m is used for geocoding. The simulated terrain phase is used for differential processing of the interferometric phase, and polynomial fitting is used for flat ground phase removal. In addition, to suppress the influence of noise, ADF or non-local method can be used for interferometric filtering.

[0093] S2-4, TCP point target

[0094] In the present invention, coherence estimation is used to measure the similarity of the surface scattering characteristics of the pixels with the same name in the primary and secondary images. Low-coherence interference patterns and point targets are eliminated to suppress incoherence noise. Different from the commonly used coherent point selection method, the present invention selects temporary coherent points as point targets, that is, the pixels can maintain high coherence within a certain observation time period. This can greatly improve the point selection density. In addition, the TCP-InSAR method uses adjacent TCP points to form arc segments to solve terrain residuals and deformation rates. This method can avoid the influence of unwrapping errors. By constructing the phase ambiguity of the Delaunay triangulation arc segment, the arc segments with integer ambiguity are eliminated. The TCP points and arc segments finally retained are used as the observation values ​​of TCP-InSAR.

[0095] S2-5, TCP-InSAR deformation estimation

[0096] For the retained TCP point targets and arc segments, the present invention constructs a deformation model using a priori models or unconstrained linear deformation models (linear rates between SAR images are not equal). In addition to deformation parameters, the model also includes terrain residuals. In TCP-InSAR, the phase difference between two adjacent TCP points is used as the observation equation of the arc segment. Considering that the atmospheric delay has a strong spatial correlation, the atmospheric phase between adjacent TCP points can be ignored after separation by differential means. In addition, the random noise after differential satisfies the statistical property of mathematical expectation being 0. For short arc segments without phase ambiguity, the least squares method can be used to estimate the parameters to be determined in the time series according to the interference pair combination method. The reference points can be used to restore the deformation time series and terrain residuals of the reservoir bank landslide.

[0097] Furthermore, in this embodiment, a combination strategy of interference patterns with multiple main images and short spatiotemporal baselines is used. Assume that there are j SAR images in a time period, and i interference patterns are selected according to the spatiotemporal baseline threshold; Assume that the deformation rate of the TCP point on the i-th interference pattern is v, and the deformation phase relative to the main image is expressed as:

[0098]

[0099] Where (l,m) is the pixel coordinate of any TCP point; is the distance between the pixel and the sensor; λ is the radar wavelength, which is related to the SAR sensor platform used, β i is the coefficient of deformation rate, combined with the interference time baseline T i Related; C i is the number of SAR images, t k is the time of the kth image, t k-1 is the time of the k-1th image, v k is the deformation rate between the two images, and v is the linear velocity vector.

[0100] T i =[t 1 -t 0 t 2 -t 1 t k -t k-1 t j -t j-1 ] (2)

[0101] The terrain residual phase is positively correlated with the interferometric spatial baseline. The DEM residual phase is expressed as:

[0102]

[0103] in, is the spatial baseline of the interferometer pair, θ i is the local angle of incidence, Δh l,m is the terrain residual, is the coefficient of the terrain residual.

[0104] TCP-InSAR uses the arc segments formed by adjacent TCP point targets as observation quantities. The phase difference between adjacent TCP points is expressed as:

[0105]

[0106] Where Δh l,m,l′,m′ is the terrain residual difference of adjacent TCP points, Δv is the linear velocity difference of adjacent TCP points, is the random noise after differentiation.

[0107] Since the atmospheric delay has a strong spatial correlation, the atmospheric phase between adjacent TCP points can be ignored after being separated by difference. Satisfies the statistical property of mathematical expectation of 0. For a short arc segment without phase ambiguity, it can be written in time series as:

[0108]

[0109] Where w is random noise, is the phase difference vector of adjacent TCP points, A = [α β] is the coefficient combination matrix of deformation parameters and DEM residuals, where β=[β 1 β 2 β i ] T The least square method can be used to solve the parameters related to Δh and Δv in equation (5). Adjacent TCP points form arc segments for parameter estimation. To improve the reliability of the solution, the effective arc segments are screened by setting the phase threshold, and the Delaunay triangulation is used as the observation of the interference pattern to re-estimate the secondary parameters. Finally, the deformation rate and DEM residual of each TCP point can be obtained by integrating the parameter estimation results relative to the reference point.

[0110] Step S3: InSAR deformation spatiotemporal evolution and time delay analysis of inducing factors

[0111] The dense monitoring points of InSAR technology provide a rich data set for analyzing the correlation between deformation signals and inducing factors. The present invention decomposes InSAR deformation into multiple components with different spatial scores and time characteristics based on independent component analysis (ICA) based on the independence assumption. The independent components decomposed by ICA can be regarded as the spatiotemporal evolution of deformation induced by different driving factors. Cross-correlation analysis is used to evaluate the correlation coefficient between deformation response and triggering factors to reveal the driving factors and time delay of reservoir bank landslides. It includes the following steps:

[0112] S3-1, Selection of the number of independent components

[0113] Reasonably determining the number of independent components is the basis of ICA independent signal separation. Too few components can easily lead to cross-fusion of different independent source signals, thereby losing the detailed features of important signals. Too many independent components will cause the signal strength of the main signal to attenuate. Therefore, the present invention uses at least two methods to estimate the number of independent components, preferably using the null space estimation method and the rank calculation method of the observation signal data matrix. Taking into account that the InSAR deformation time series will inevitably be affected by noise, the present invention uses principal component analysis (PCA) to reduce dimension and noise to select independent components. The variance ratio of each component can be used as an empirical indicator for selecting the principal component. Based on this, the final number of independent components is determined.

[0114] S3-2, ICA decomposition

[0115] ICA can decompose the mixed observation signals on the SAR time series into linear combinations of independent components with statistical characteristics. In the InSAR results, the temporal deformation of each pixel is regarded as the sum of various independent deformation signal sources and noise.

[0116] Based on independent component analysis, the present invention makes full use of the high-order statistical independence between different deformation signals and multi-source errors in InSAR time series to extract and decompose the surface deformation signals and errors caused by different driving factors. The results of these signals are expressed as linear combinations of different components and independent components including time vectors of linear deformation, periodic deformation, step deformation, atmospheric delay and noise.

[0117] In this embodiment, suppose a group of InSAR observation values ​​X 1 (t),X 2 (t),…,X n (t), t is the observation time, and the observation value is generated by the linear combination of independent components, which is:

[0118] [X 1 (t),X 2 (t),…,X n (t)] T =A.[S 1 (t),S 2 (t),…,S n (t)] T (6)

[0119] Among them, A is an unknown coefficient matrix representing the linear combination of different components; S(t) is the independent component, including linear deformation, periodic deformation, step deformation, atmospheric delay and noise error. Independent Component Analysis (ICA) is based on the independence assumption to decompose the InSAR deformation time series into multiple components with different spatial and temporal characteristics. i (t) to estimate the coefficient matrix A and independent components S i (t).

[0120] S3-3, InSAR deformation spatiotemporal evolution and correlation analysis

[0121] The principal component signals of the ICA of the reservoir bank landslide are identified as independent components related to the landslide movement. Generally, different independent components will show different spatiotemporal evolution characteristics. Usually, the spatial evolution characteristics of the reservoir bank landslide are manifested as strong deformation zones near the reservoir such as the front edge and toe of the landslide. The spatial vector is inversely proportional to the distance from the reservoir, which may be driven by reservoir water level fluctuations. In addition, strong deformation zones may also appear in the middle and rear edges of the landslide due to the influence of rainfall infiltration. In terms of time evolution, the step and periodic fluctuations of the front edge of the landslide and the linear creep of the rear edge of the landslide are all common movement characteristics. To this end, the present invention combines the spatial evolution to conduct a correlation analysis between the time vector of the independent component and the reservoir water level fluctuation, accumulated rainfall and the joint effect of reservoir water, so as to reveal the potential driving factors.

[0122] In the present invention, FastICA is used to process InSAR time series. First, the central observation matrix X is calculated according to the observation matrix X. C :

[0123]

[0124] in, Each column of is equal to the average of the corresponding column in X, and the center matrix X C The mean of each column of is zero. In order to perform whitening, the present invention uses the center matrix to generate the covariance matrix C X for:

[0125]

[0126] The covariance matrix is ​​decomposed by maximizing the total variance of the projection based on principal component analysis (PCA):

[0127]

[0128] Among them, e and E are eigenvectors and eigenvector matrices, d and D are eigenvalues ​​and eigenvalue matrices respectively. The calculation formula of the whitening matrix is:

[0129] Q=(ED -1 / 2 E T ) k (10)

[0130] Among them, k is the number of principal components retained, and the observation matrix Z after whitening is written as:

[0131] Z=QX C (11)

[0132] After the observation matrix is ​​centered and whitened, the FastICA algorithm estimates the source signal matrix S and the mixing matrix A by maximizing the spatial non-Gaussianity.

[0133]

[0134] in, and are the mixing matrix and source matrix obtained from the center and whitened observations respectively. By using the inverse operation of the whitening process, the center observation matrix is ​​expressed as a combination of the source matrix and the mixing matrix:

[0135]

[0136] Among them, Q + is the pseudo-inverse of the whitening matrix Q. The original observation matrix is:

[0137]

[0138] By performing FastICA processing on the TCP-InSAR observation results, the time eigenvectors and spatial scores of the independent components are separated. Due to the temporal and spatial differences of triggering factors such as rainfall and reservoir water level fluctuations, the movement characteristics of the reservoir bank landslide will show a variety of forms. The time eigenvectors and spatial scores separated by ICA can be regarded as the temporal and spatial deformation evolution caused by different triggering factors. Among them, the spatial distribution of the triggering factors can be identified from the spatial scores. The present invention uses the normalized independent component eigenvectors for cross-correlation analysis with possible driving factors, which include rainfall, periodic water level fluctuations, and the superposition effect of rainfall and water level fluctuations. In order to quantify the impact of the corresponding triggering factors, the present invention uses cross-correlation analysis to estimate the lag time of the deformation, and takes the step size with the largest correlation as the lag time of the triggering factor.

[0139] Step S4: Fluid-Structure Interaction Numerical Model

[0140] With the ability to solve complex geotechnical engineering problems by multi-field coupling, numerical simulation methods have gradually become an important means of evaluating the stability of large-scale reservoir bank landslides. In terms of numerical calculation, Geostudio based on finite elements and FLAC3D based on finite differences are two commercial software commonly used for numerical simulation of reservoir bank landslides at this stage. GeoStudio has powerful saturated-unsaturated seepage calculation functions, can set complex hydraulic boundary conditions, and has built-in multiple limit equilibrium methods. It is often used for seepage field simulation and stability analysis of reservoir bank landslides. However, GeoStudio has few built-in constitutive models, secondary development is difficult, and it cannot simulate large deformations. Therefore, this method is difficult to use for the deformation and failure mechanism analysis of reservoir bank landslides. FLAC3D has rich built-in elastic and plastic material constitutive models, diverse structural forms, easy secondary development, and good convergence effect of large deformation simulation. It is often used for deformation and failure mechanism analysis of landslides. However, it is difficult for FLAC3D to perform transient seepage analysis of unsaturated soils, which limits the application of this method in reservoir bank landslides.

[0141] In view of the limitations of the two numerical calculation methods in the application of reservoir bank landslide, this paper will establish a reservoir bank landslide multi-field information flow-solid coupling numerical model that can be quickly converted between GeoStudio and FLAC3D to perform flow stability analysis of reservoir bank landslide under rainfall and water level fluctuation conditions. It includes the following steps:

[0142] S4-1, InSAR deformation inversion saturated permeability

[0143] In reservoir bank landslide, the rationality of numerical simulation results is easily restricted by the accuracy of saturated permeability coefficient. Due to the existence of spatial heterogeneity and disturbance of sample data, it is very difficult to accurately determine the saturated permeability.

[0144] The present invention inverts the hydraulic diffusion coefficient based on the rainfall-induced temporal deformation decomposed by ICA and combined with daily rainfall records, obtains the saturated permeability estimation according to the relationship between the hydraulic diffusion coefficient and the permeability coefficient, and estimates the soil-water characteristic curve and the permeability coefficient function in combination with geotechnical test results.

[0145] In this embodiment, the estimation of permeability coefficient is crucial in the seepage analysis. However, the permeability coefficient measured experimentally is usually within a range of values, and the magnitude difference of this range can be as high as 1.0e -2 The one-dimensional pore water diffusion model can be used to invert the pore water diffusion coefficient from InSAR time series deformation and rainfall records.

[0146] In the present invention, this model is used to represent the change of transient pore water pressure caused by rainfall infiltration:

[0147]

[0148] Where P is the pore water pressure, D is the hydraulic diffusion coefficient, and z is the depth. This model describes the propagation of pore water pressure inside the landslide. The pore water pressure on the surface is described by the rainfall record R(t) and the scale factor q:

[0149] P(t,Z=0)=qR(t) (16)

[0150] The analytical solution of this propagation process is:

[0151]

[0152] Where s is the time variable. The hydraulic diffusion coefficient and permeability coefficient are combined as follows:

[0153]

[0154] Where C is a measure of the change in volumetric water content with pore pressure, and takes the minimum value C when the soil is saturated. 0 , the hydraulic diffusion coefficient will reach a peak value D 0 , k s represents the saturated permeability coefficient.

[0155] S4-2, Geostudio unsaturated seepage field

[0156] The engineering geological profile obtained from geological survey can be used as the basis for material partitioning and establishing numerical models. When establishing node and grid models, the geometric size and calculation precision of the reservoir bank landslide are taken into consideration. The sliding zone and sliding body are usually divided into high-resolution grids, while the grid spacing of the bedrock layer can be appropriately relaxed. Hydraulic boundary conditions are indispensable for solving the seepage field.

[0157] Considering that the dynamic seepage of the reservoir bank landslide can be affected by both rainfall and reservoir water level, the hydraulic boundary conditions are set in the Geostudio unsaturated seepage analysis as follows:

[0158] ① In order to observe the influence of the transient saturated zone formed by rainfall infiltration on the deformation and stability of the landslide, the present invention sets the rear edge of the landslide as a fixed total water head boundary condition;

[0159] ② Set the unit flow boundary condition on the landslide surface, and the unit flow is equal to the rainfall intensity. It should be noted that when the rainfall intensity exceeds the saturated permeability coefficient of the sliding material, the slope surface is converted to a zero pressure boundary condition;

[0160] ③ The landslide slope is set as the total head boundary condition below the normal water level of the reservoir, and the height is equal to the reservoir water level. It should be noted that when the water level fluctuates, the slope from the reservoir water level to the normal water level will be set with drainage conditions, that is, there may be overflow points of groundwater inside the slope;

[0161] ④ Impermeable boundary conditions are set above the infiltration line of the rear edge of the landslide, the bottom of the model, and the side of the model reservoir.

[0162] Based on this, a numerical model for unsaturated seepage analysis of reservoir bank landslides in Geostudio was established. Figure 2 As shown in Figure 1, the numerical model of a typical landslide in a reservoir area in a certain region.

[0163] In this embodiment, the rapid rise and fall of the reservoir water level and the infiltration of rainfall will change the original mechanical state and shear strength of the geotechnical materials. Therefore, the change of the seepage field should be considered first in the numerical simulation of the reservoir bank landslide. The present invention establishes a fluid-solid analysis numerical model coupled with pore pressure based on the engineering geological profile of a typical landslide to analyze the deformation mechanism, failure mode and safety factor dynamic response law of the landslide.

[0164] In the seepage field, the general governing differential equation of two-dimensional seepage is used to describe the difference in the flow of fluid into and out of the unit cell at a certain time at the node in the model. This difference is equal to the change in soil volume water content:

[0165]

[0166] Where H is the total hydraulic head, k is the permeability coefficient, and represents the x and y directions, respectively. Q is the applied boundary flow, θ is the volumetric water content, and t is the time.

[0167] In steady-state seepage, the right-hand side of equation (19) is equal to 0. In reservoir bank landslides, rainfall infiltration and water level fluctuations will cause changes in the saturation of the rock and soil mass, and this change shows that the rock and soil mass has unsaturated characteristics. In view of the seepage problem of unsaturated soil, the present invention estimates SWCC based on the moisture content of the material sample function in the Geostudio SEEP / W module. Then, HCF is estimated based on SWCC by the Van Genuchten method.

[0168] S4-3, fast conversion between Geostudio and FLAC3D

[0169] Since the finite element method uses two-dimensional seepage analysis, the types of grid elements are triangles and quadrilaterals. However, FLAC3D uses a three-dimensional scene.

[0170] According to the node numbering rules of different grid types of two-dimensional model and three-dimensional model, a corresponding relationship is established. Based on this corresponding relationship, the present invention converts the pore water pressure, saturation and volume water content in the finite element seepage field into the FLAC3D model in a 1:1 manner. This conversion depends on the unit node ID, coordinates, pore pressure, saturation and volume water content in the seepage field. Figure 3 The before and after scenarios of pore water pressure and saturation in the initial state are shown.

[0171] According to the effective stress principle, changes in seepage conditions will inevitably affect the mechanical properties of the soil. Therefore, many researchers conduct landslide stress analysis by coupling the seepage field. Among them, Geostudio has built-in multiple limit equilibrium methods, which are often used to calculate the stability coefficient of reservoir bank landslides. However, GeoStudio cannot simulate large deformations. The FLAC3D software developed based on the fast Lagrangian finite difference method overcomes the disadvantage of large deformations being difficult to converge, but it is difficult to set complex hydraulic boundary conditions for unsaturated seepage analysis.

[0172] In the present invention, the seepage field calculated by Geostudio is coupled to the FLAC3D model to analyze the deformation and failure mechanism of the reservoir bank landslide. The conversion method is as follows: Figure 3 In this embodiment, the model conversion of the seepage field is to stretch the two-dimensional grid model of Geostudio in the normal direction. In the three-dimensional model after stretching, the original triangle and quadrilateral units are transformed into wedge-shaped units and hexagonal block units. Figure 3 The node correspondence, pore pressure and saturation before and after the model is stretched are shown.

[0173] S4-4, Initial stress and parameter correction

[0174] When the pore water pressure is imported into the FLAC3D model, the ratio of the pore water pressure to the effective stress in the initial total stress of the model will change. However, the initial total stress in the model does not change, which will cause the initial effective stress of the model after the pore pressure is imported to be inconsistent with the initial effective stress before the import.

[0175] The present invention extracts seepage information in the fluid-solid coupling model based on the effective stress principle and the relationship between pore water pressure, matrix suction, total stress, saturation, etc. to realize the initial stress correction of saturated and unsaturated zones. While rainfall infiltration and water level fluctuations cause changes in pore pressure, they also cause changes in the weight and strength of geotechnical materials. The changes in these parameters are closely related to the changes in volumetric water content (saturation). Therefore, the present invention performs zoning corrections on geotechnical material parameters based on the results of the seepage field. The zoning correction implementation process is similar to the initial stress correction.

[0176] S4-4-1, Correction of Initial Stress

[0177] When the pore pressure is introduced into the FLAC3D model, the magnitude of the initial effective stress in the model will change. The present invention corrects the initial total stress after the pore pressure is introduced. According to the effective stress principle, the effective stress σ′ of the rock mass can be expressed as:

[0178]

[0179] Where σ is the total stress, u a is the pore gas pressure, which is generally taken as 0; σ s is the suction stress, which is equal to the pore pressure when the soil is saturated and is taken as the matrix suction when the soil is unsaturated. r is the saturation. The suction stress and total stress of the model before the pore water pressure is introduced are set to be σ s 0 and σ 0 , the suction stress of the model after import is σ s 1 , the corrected initial total stress σ 1 It can be expressed as:

[0180] σ 1 =σ 0 -(σ s 0 -σ s 1 ) (twenty one)

[0181] S4-4-2, Modification of model parameters

[0182] Let ρ d is the dry density of the material. The present invention uses the porosity n and the water density ρ w Correction for the natural density ρ:

[0183] ρ=ρ d +nS r ρ w (twenty two)

[0184] The transient saturated zone formed by rainfall infiltration will reduce the shear strength of the rock and soil. For the unsaturated zone, the present invention uses pore water pressure and friction angle to correct the effective cohesion:

[0185]

[0186] Where C is the cohesion under saturation state.

[0187] Step S5: Deformation mechanism of reservoir bank landslide

[0188] like Figure 4 and Figure 5 As shown in Figure 3, the InSAR deformation results separated by ICA describe the temporal and spatial evolution characteristics of the reservoir bank landslide and the potential inducing factors, but it is difficult to reveal the mechanical mechanism and stability of the deformation.

[0189] The present invention uses the driving factors of InSAR deformation as hydraulic boundary conditions to perform stress-strain analysis. The failure deformation mode, stress-strain state and stability of different inducing factors are analyzed. The following steps are included:

[0190] S5-1, Stress Boundary Conditions

[0191] The present invention sets stress or displacement rate boundary conditions on the left and right sides, bottom, slope, front and back of the model to constrain the numerical model. The detailed settings are as follows:

[0192] ①Fix the normal displacement rate on the left and right sides of the model, that is, velocity-x=0;

[0193] ② Set the normal displacement rate of the front and rear of the model to 0;

[0194] ③ At the bottom of the model, fix all displacement rates of the model, i.e. velocity-x and velocity-y = 0;

[0195] ④ Below the normal water level of the reservoir slope, a hydrostatic pressure boundary condition corresponding to the slope water level height is set to simulate the back pressure effect of the water level on the slope surface. Figure 2 The stress boundary condition setting of a typical landslide in a reservoir area in a certain region is demonstrated.

[0196] S5-2, Calculation Condition Settings

[0197] The external dynamic factors that affect the stability of the reservoir bank landslide include rainfall and reservoir water level fluctuations. According to the different stages of the landslide and the different conditions of the external dynamics, the present invention designs a variety of calculation conditions to observe the stress-strain characteristics of the reservoir bank landslide. The main calculation conditions are set as follows:

[0198] (1) Only the impact of rainfall infiltration is considered, and the reservoir water level remains at normal level;

[0199] (2) Only the change of pore water pressure in the landslide caused by water level fluctuation is considered, and the rainfall intensity on the slope is set to 0;

[0200] (3) Considering the coupling effect of rainfall and reservoir water level to observe the stress-strain and stability characteristics of the landslide;

[0201] (4) Consider the stress-strain characteristics of the landslide under different water level drop or rise rates.

[0202] The mechanical mechanism of landslide deformation is assisted by the stability, destruction deformation characteristics and mechanical state in different scenarios.

[0203] S5-3, Dynamic response of reservoir bank landslide stability during InSAR observation period

[0204] Generally, the reduction of landslide stability will induce deformation. The present invention uses the water level fluctuation and daily rainfall intensity during the InSAR observation period as the calculation conditions to observe the spatiotemporal evolution of the simulated deformation and the dynamic response law of stability. The correlation analysis of the independent components of the InSAR deformation can clearly show the response law of the dynamic stability coefficient to rainfall and reservoir water level fluctuation. In addition, affected by the low permeability coefficient, this response can also reveal the time delay of stability.

[0205] Based on the coupling model, the strength reduction method is used to quantitatively solve the failure deformation model, stress-strain state and stability of the model to analyze the deformation mechanism of the reservoir bank landslide.

[0206] The above implementation cases are only preferred embodiments of the present invention and are not intended to limit the present invention in any form. Although the present invention has been disclosed as above with preferred embodiments, they are not intended to limit the present invention. Therefore, any simple modification, equivalent changes and modifications made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for analyzing the deformation mechanism of reservoir bank landslides by combining InSAR deformation characteristics and fluid-solid coupling, characterized in that: The analysis method first uses multi-temporal InSAR to restore the deformation field of the reservoir bank landslide and combines the ICA method to analyze the inducing factors of the independent components of the deformation; then, the saturated permeability coefficient of the sliding material is corrected based on the InSAR deformation time series; then, the seepage analysis results are coupled to the FLAC3D model for stress-strain analysis; finally, the deformation mechanism of the reservoir bank landslide is revealed based on the InSAR spatiotemporal evolution characteristics and fluid-solid coupling analysis results.

2. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 1 is characterized in that: The analytical method comprises the following steps: Step S1, acquiring SAR, hydrological and geological data: hydrological data is used to analyze InSAR deformation driving factors and time delays and as boundary conditions for numerical simulations, and geological data is used to establish numerical models; Step S2, TCP-InSAR data processing: using the temporary coherence point InSAR method to restore the InSAR deformation field of the reservoir bank landslide; Step S3, InSAR deformation spatiotemporal evolution and time delay analysis of inducing factors: Independent component analysis based on the independence assumption decomposes the InSAR deformation into multiple components with different spatial scores and time characteristics; the decomposed independent components are used as the spatiotemporal evolution of deformation induced by different driving factors, and the correlation coefficient between the deformation response and the triggering factor is evaluated by cross-correlation analysis to analyze the driving factors and time delay of the reservoir bank landslide; Step S4, fluid-solid coupling numerical model: establish a multi-field information fluid-solid coupling numerical model of the reservoir bank landslide that can be quickly converted between GeoStudio and FLAC3D, and conduct flow stability analysis of the reservoir bank landslide under conditions of rainfall and water level fluctuation; Step S5, deformation mechanism of reservoir bank landslide: using the driving factors of InSAR deformation as hydraulic boundary conditions, stress-strain analysis is performed to analyze the failure deformation mode, stress-strain state and stability of different inducing factors.

3. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 2 is characterized in that: The step S3 includes the following steps: S3-1, select the number of independent components: use principal component analysis (PCA) to reduce dimension and noise to select independent components, and the variance ratio of each component is used as an empirical indicator for selecting principal components; S3-2, ICA decomposition: ICA decomposes the mixed observation signals on the SAR time series into linear combinations of independent components with statistical characteristics; the temporal deformation of each pixel is the sum of independent deformation signal sources and noise; the surface deformation signals and errors caused by different driving factors are extracted and decomposed; S3-3, InSAR deformation spatiotemporal evolution and correlation analysis: Combined with spatial evolution, the time vector of the independent component is correlated with reservoir water level fluctuation, accumulated rainfall and reservoir water joint effect to analyze the potential driving factors.

4. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 3 is characterized in that: In the ICA decomposition, suppose a set of InSAR observations X1(t), X2(t),…, X n (t), t is the observation time, and the observation value is generated by the linear combination of independent components, which is: [X1(t),X2(t),…,X n (t)] T =A.[S1(t),S2(t),…,S n (t)] T (6) Among them, A is an unknown coefficient matrix representing the linear combination of different components; S(t) is an independent component, including linear deformation, periodic deformation, step deformation, atmospheric delay and noise error; when X is known, i (t) to estimate the coefficient matrix A and independent components S i (t).

5. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 4 is characterized in that: In the spatiotemporal evolution of InSAR deformation, FastICA is used to process the InSAR time series. First, the central observation matrix X is calculated according to the observation matrix X. C : in, Each column of is equal to the average of the corresponding column in X, and the center matrix X C Each column of has a mean of zero; the covariance matrix C is generated using the center matrix X for: The covariance matrix is ​​decomposed by maximizing the total variance of the projection based on principal component analysis: Among them, w and E are eigenvectors and eigenvector matrices, d and D are eigenvalues ​​and eigenvalue matrices respectively; the calculation formula of the whitening matrix is: Q=(ED -1 / 2 AND T ) k (10) Among them, k is the number of principal components retained, and the observation matrix Z after whitening is written as: Z=QX C (11) After the observation matrix is ​​centered and whitened, the FastICA algorithm estimates the source signal matrix S and the mixing matrix A by maximizing the spatial non-Gaussianity: in, and are the mixing matrix and source matrix obtained from the center and whitened observations respectively. By using the inverse operation of the whitening process, the center observation matrix is ​​expressed as a combination of the source matrix and the mixing matrix: Among them, Q + is the pseudo-inverse of the whitening matrix Q; the original observation matrix is: By performing FastICA processing on TCP-InSAR observations, the temporal eigenvectors and spatial scores of independent components are separated.

6. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 3 is characterized in that: The step S4 includes the following steps: S4-1, InSAR deformation inversion of saturated permeability: Invert the hydraulic diffusion coefficient based on the rainfall-induced temporal deformation decomposed by ICA and the daily rainfall records, and obtain the saturated permeability estimate based on the relationship between the hydraulic diffusion coefficient and the permeability coefficient; estimate the soil-water characteristic curve and the permeability coefficient function in combination with the geotechnical test results; S4-2, Geostudio unsaturated seepage field: hydraulic boundary conditions were set in Geostudio unsaturated seepage analysis, and a numerical model of Geostudio unsaturated seepage analysis of reservoir bank landslide was established; S4-3, quick conversion between Geostudio and FLAC3D: the finite element mesh model is stretched in the normal direction, and the original triangle and quadrilateral units in the stretched three-dimensional model are changed into wedge and hexagonal block units; the corresponding relationship is established according to the node numbering rules of different mesh types of the two-dimensional model and the three-dimensional model; the pore water pressure, saturation and volume water content in the finite element seepage field are converted to the FLAC3D model in a 1:1 ratio; S4-4, initial stress and parameter correction: In the fluid-solid coupling model, based on the effective stress principle and the relationship between pore water pressure, matrix suction, total stress and saturation, the seepage information is extracted to correct the initial stress of the saturated and unsaturated zones; and the geotechnical material parameters are zoned and corrected based on the results of the seepage field.

7. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 6 is characterized in that: The inversion process of the hydraulic diffusion coefficient is: The formula for expressing the transient pore water pressure change caused by rainfall infiltration is: Where P is the pore water pressure, D is the hydraulic diffusion coefficient, and z is the depth. Formula (15) describes the propagation process of pore water pressure inside the landslide. The rainfall record R(t) and the scale factor q are used to describe the pore water pressure on the surface: P(t,z=0)=qR(t) (16) The analytical solution of the propagation process is: Where s is the time variable; the hydraulic diffusion coefficient and permeability coefficient are combined as follows: Where C is a measure of the change in volumetric water content with pore pressure. It takes the minimum value C0 when the soil is saturated, and the hydraulic diffusion coefficient reaches the peak value D0. s represents the saturated permeability coefficient.

8. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 6 is characterized in that: The correction method for the initial stress is: According to the effective stress principle, the effective stress σ′ of rock and soil mass is expressed as: Where σ is the total stress, u a is the pore gas pressure; σ s is the suction stress, which is equal to the pore pressure when the soil is saturated and is taken as the matrix suction when the soil is unsaturated. r is saturation; The suction stress and total stress of the model before the pore water pressure is introduced are set to be σ s 0 and σ0, the suction stress of the model after import is σ s 1, the corrected initial total stress σ1 is expressed as: σ1=σ0-(σ s 0-s s 1) (21) 9. The method for analyzing the deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 8 is characterized in that: The correction method of the parameters is: Let ρ d is the dry density of the material, using the porosity n and the density of water ρ w Correction for the natural density ρ: p=p d +nS r r w (22) For the unsaturated zone, the effective cohesion is modified using the pore water pressure and the friction angle: Where C is the cohesion under saturation state.

10. The method for analyzing deformation mechanism of reservoir bank landslide combining InSAR deformation characteristics and fluid-solid coupling according to claim 3 is characterized in that: The step S5 includes the following steps: S5-1, stress boundary conditions: setting boundary conditions to constrain the numerical model; S5-2, calculation condition setting: the calculation condition setting is carried out according to different stages of the landslide and different conditions of external dynamics, and the mechanical mechanism of landslide deformation is analyzed based on the stability, damage deformation characteristics and mechanical state under different scenarios; S5-3, Dynamic response of reservoir bank landslide stability during InSAR observation period: The water level fluctuation and daily rainfall intensity during the InSAR observation period were used as calculation conditions to observe the spatiotemporal evolution of simulated deformation and the dynamic response of stability.

Images